IRQ #0 broken on ARM

[Dmitry Eremin-Solenikov]

2014-11-21 13:52 GMT+03:00 Marc Zyngier <marc.zyngier at arm.com>:
> On Fri, Nov 21 2014 at 10:31:05 am GMT, Dmitry Eremin-Solenikov <dbaryshkov at gmail.com> wrote:
>> Hello,
>>
>> After the commit a71b092a9c68685a270ebdde7b5986ba8787e575
>> (ARM: Convert handle_IRQ to use __handle_domain_irq) IRQ #0 is broken
>> on ARM. It is a valid IRQ and it is quite imporant (on sa1100 it's a GPIO0).
>
> Well, this is a valid IRQ number if you're not using irq domains. I may
> be a bit pedantic here, but I thing this is an important distinction.
>
>> The worst thing is that the CPU will be stuck busy-looping around this
>> IRQ w/o printing anything to the console or masking the irq. How
>> should we cope with that? I'd like to propose to either revert the
>> offending commit or to add the following patch.
>
> Well, said commit fixes a rather important bug, so I suggest we keep
> around. Now, as for your suggestion:

>From the commit message it was not clear, that there was a bug fixed
(it talks only about code duplication).

[skipped]

> As I mentioned above, IRQ0 is not valid when using irq domains. As an
> alternative, how about this:
>
> diff --git a/kernel/irq/irqdesc.c b/kernel/irq/irqdesc.c
> index a1782f8..9f5bc92 100644
> --- a/kernel/irq/irqdesc.c
> +++ b/kernel/irq/irqdesc.c
> @@ -365,7 +365,7 @@ int __handle_domain_irq(struct irq_domain *domain, unsigned int hwirq,
>          * Some hardware gives randomly wrong interrupts.  Rather
>          * than crashing, do something sensible.
>          */
> -       if (unlikely(!irq || irq >= nr_irqs)) {
> +       if (unlikely((lookup && !irq) || irq >= nr_irqs)) {
>                 ack_bad_irq(irq);
>                 ret = -EINVAL;
>         } else {
>
> I don't have a platform to test this on, but maybe you could give it a
> go and let me know if that helps?

It helps in my case. Thank you. Please add me to Cc if you submit
this patch.

-- 
With best wishes
Dmitry